State-of-the-art video coding schemes, such as the upcoming HEVC (High-Efficiency Video Coding), supports splitting a picture into smaller rectangular units called tiles units. Each tile units can be independently encoded and decoded by separate encoders and decoders, respectively. The primary purpose of tiles units is to allow parallel processing of the picture to reduce implementation cost and complexity.
The method to split a picture into smaller tiles in the prior art is to cut the picture horizontally into columns first and then vertically into rows. This method of splitting the picture is shown in FIG. 13. By splitting the picture in this manner, transmitting all the column widths and then all the row heights will be sufficient to define the tiles region in a picture.
In the tiles coding scheme used in prior art, the tiles units are coded separately and the spatial dependency between tiles can be removed. However, the inter prediction of the tiles can go across the tiles region in the reference pictures.